Optimization MCP

# Changelog All notable changes to this project will be documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). ## [Unreleased] ### Planned for v2.6.0 - Monte Carlo integration patterns documentation - Schema Grammar reference section - Quick Start in 60 seconds guide - Example JSON payloads in docs/examples/ ## [2.5.1] - 2025-12-06 ### Fixed - Cross-platform path handling using pathlib (Windows + Unix compatible) - Item universe consistency validation (prevents typo bugs in item names) - Removed misleading test claims from README (no public test directory) ### Changed - Logging now configurable via `OPT_MCP_LOG_PATH` environment variable - Logging includes StreamHandler for stdout visibility - Requirements pinned to ranges (>=X,<Y) for better reproducibility ### Added - Comprehensive scalability guidelines for all 9 tools - Performance characteristics documentation - "When You Hit Scale Limits" guidance section ### Documentation - Removed test badge and test count claims from README - Added scale limits for all tools (items, resources, time horizons) - Changed "Battle-Tested: 51/51 passing" to "Production Ready: Internally tested" ## [2.5.0] - 2025-12-04 ### Added - `optimize_network_flow` tool - Network flow optimization with NetworkX (5000x speedup) - `optimize_pareto` tool - Pareto frontier generation for multi-objective - `optimize_stochastic` tool - 2-stage stochastic programming - `optimize_column_gen` tool - Column generation framework - NetworkXSolver backend - Specialized network flow algorithms - Network validation in DataConverter - Pareto frontier visualization and sketchnote ### Performance - Network flow: 5000x faster than general LP for pure network problems - All tools handle 1K-10K variable scale efficiently - Pareto frontier: 20-100 points in 10-30 seconds ### Testing - 17 new tests added (8 network flow + 6 Pareto + 3 stochastic) - Total: 51/51 tests passing (internal) ## [2.4.0] - 2025-12-04 ### Added - `optimize_stochastic` tool - 2-stage stochastic programming - Extensive form implementation for scenario-based decisions - Expected value and worst-case risk measures - VSS (Value of Stochastic Solution) calculations - 3 comprehensive tests ### Features - Sequential decision modeling (order now, adjust later) - Scenario-based uncertainty representation - Recourse decisions after information revelation ## [2.3.0] - 2025-12-04 ### Added - `optimize_pareto` tool - Pareto frontier generation - Weighted sum scalarization for multi-objective - Knee point recommendation algorithm - Trade-off analysis between objectives - Mixed sense support (maximize + minimize simultaneously) - Dominated solution filtering - 6 comprehensive tests ### Features - Generate 20-100 non-dominated solutions - Explore full trade-off space - Identify best-balanced solution automatically ## [2.2.0] - 2025-12-04 ### Added - `optimize_network_flow` tool - Network flow optimization - NetworkXSolver class - Specialized network flow backend - Network structure validation in DataConverter - Support for min-cost flow, max-flow, assignment problems - Bottleneck analysis for capacity-constrained edges - 8 comprehensive tests - networkx>=3.0 dependency ### Performance - 5000x+ speedup for network flow problems vs general LP - Small networks (<100 nodes): <0.1 seconds - Large networks (1000-10000 nodes): <10 seconds ### Technical - Exploits network structure (incidence matrix) - Network simplex algorithm for min-cost flow - Edmonds-Karp algorithm for max-flow ## [2.1.0] - 2025-12-02 ### Added - `optimize_execute` tool - Custom optimization with auto-solver selection - Flexible dict-based problem specification - Auto-detection of problem type (LP/MIP/QP/nonlinear) - Solver override capability - 5 comprehensive tests ### Features - Power user tool for custom formulations - Escape hatch for problems not fitting standard templates - Auto-selects best solver (PuLP/SciPy/CVXPY) - Flexible variable types (continuous, integer, binary) ## [2.0.0] - 2025-12-02 ### Added - Multi-objective optimization (weighted scalarization) - CVXPYSolver for quadratic programming - `optimize_portfolio` tool (Sharpe, variance, return objectives) - `optimize_schedule` tool (dependencies, makespan, critical path) - Enhanced constraints (conditional, disjunctive, mutex) - Risk contribution analysis for portfolios - Critical path detection for schedules - 26 new tests - Comprehensive documentation ### Changed - `optimize_allocation` now supports multi-objective via "functions" key - CVXPY solver selection uses SCS for robustness - README reorganized with 4 tool sections ### Features - Balance multiple competing objectives - Portfolio risk-return optimization - Project scheduling with resource constraints - If-then, OR, XOR constraint logic ## [1.0.0] - 2025-12-01 ### Added - Initial release - `optimize_allocation` tool - Resource allocation - `optimize_robust` tool - Robust optimization - PuLP and SciPy solvers - Monte Carlo integration (3 modes: percentile, expected, scenarios) - Basic test suite - Installation and usage documentation ### Features - Resource allocation under constraints - Robust optimization across scenarios - Multi-resource support - Shadow price analysis --- ## Version Numbering - **Major (X.0.0)**: Breaking API changes - **Minor (0.X.0)**: New features, backward compatible - **Patch (0.0.X)**: Bug fixes, documentation ## Links - [Full Documentation](README.md) - [Installation Guide](docs/INSTALLATION.md) - [GitHub Repository](https://github.com/eesb99/optimization-mcp)